Research On System Of Fish Tracking And Movement Behavior Modeling For Water Quality Monitor

Biomonitoring techniques, with the vantage of biological basis for water quality assessment by observing physiological properties and different reactions of aquatic animals to water qualities and pollution degree, have been universally applied to water quality monitor and security pre-warning systems of environmental quality. Compared to traditional physico-chemical analysis, biological detection is capable of analyzing the interplays of multiple toxic substances and determining the direct relations between the mass concentration of toxic substances and movement features of aquatic animals. Thus, the attention of automatically biological monitoring of water quality is concentrated on effective and efficient extractions of the movement features of aquatic animals.By taking fishes as the sensors of water quality monitor, this paper has analyzed the application of computer vision in water quality monitoring, and a system of fish tracking and movement behavior modeling for water quality monitor has been constructed initially in the process. The research focuses of this paper include real-time detecting of the fish movement, constructions of both the behavioral modeling and the monitoring platform of fish movement. To be specific, followings are four focal perspectives to be expounded in this paper.The real time detecting method based on fuzzy inference of background difference is first to be explored in this paper. In the attempt to improve the refresh rate and quality of the background frame offered by traditional background difference algorithm, this paper has proposed another algorithm for updating background frame on the basis of fuzzy inference and difference in frame. Through fuzzy inference, the background of the object will be extracted efficiently and the foregrounds will be correctly separated in real time. Meanwhile, fuzzy inference of anti-noises will also be introduced to strengthen the robustness of the algorithm and further to overcome environmental factors.Second, the object tracking algorithm based on twice searching of auto-Camshift will be explored. As to the fact that traditional Camshift algorithm fails to realize auto and multi-objects tracking, this paper, with fuzzy inference background difference and twice searching as its major methodology, has proposed the auto-Camshift tracking algorithm to feasibly overcome traditional tracking problems such as incompetent of auto-tracking and unsatisfactory tracking results. Furthermore, contour marks and multi-Camshift tracker are introduced to achieve the multi-object tracking of Camshift algorithm.Third, this paper has demonstrated the PTW based modeling method of fish movement behavior, which will be expounded on the basis of displacement of the fish movement tracked and assessed as well as the velocity and angle velocity calculated. Meanwhile, the velocity and locus model of fish movement under normal condition is also established to offer a criterion for abnormal fish movement.Last but not least, a system platform of fish tracking and movement behavior modeling for water quality monitor has also been proposed and a software platform for fish video capturing model, vision processing model and fish behavioral model initially constructed.